Generative AI as a Catalyst for Data-Driven Learning: Efficacy, Equity, and Engagement in Translation Education

Generative artificial intelligence (AI) tools like DeepSeek hold transformative potential for data-driven learning (DDL) in translator education, yet their efficacy, equity implications, and impact on learner engagement remain underexplored, particularly outside Western-dominated research. Grounded in the Translatorial Competence Model and Cognitive Load Theory, this mixed-methods study investigated the integration of generative AI into DDL activities for 80 translation majors from a comprehensive university in eastern China over a 12-week intervention. Results indicated that AI-enhanced DDL was associated with significantly improved terminology mastery and cultural adaptation compared to traditional methods. The experimental group showed greater gains in accuracy (23% vs. 14%), terminology consistency (28% vs. 17%), and cultural appropriateness (19% vs. 9%). However, efficacy was not uniform. A significant rural-urban divide mediated outcomes, with urban students with prior AI exposure outperforming rural peers (β = .32, p = .008). This disparity was significantly reduced through structured pedagogical scaffolding, with the rural-urban proficiency gap narrowing from 14% to 5%. Learners also reported heightened motivation ( M = 4.2 vs. 3.5 control, p < .001) but raised concerns about AI’s potential for errors and Western-centric biases. The study underscores that the integration of generative AI is highly conditional, requiring targeted strategies to address socioeconomic equity gaps and ensure cultural relevance. These findings offer a cautious framework for integrating AI into translator education, highlighting the prerequisites for equitable, discovery-based learning in diverse contexts.